An AGARCH Model with Constraints

Since the first ARCH model introduced by Engle （1982）, its various extensions keep emerging one after another. However, there is a common backward among these models: only nonnegativity is supposed on the coeffcients in their conditional variance functions to guarantee conditional variance nonnegative,and the maxi- mum likelihood method is used to estimate the models. This paper suggests a new concept of AGARCH model with constraints,and sue nonlinear programming method to estimate the model instead of MLE.The empirical study shows that the model suggested in this paper is feasible and better than the traditional ones.